548 results about "Visual task" patented technology

The invention provides a visual project management method and system based on task tracking, and relates to the technical field of project management, and the method comprises the steps: collecting project risk information through employing a context awareness technology, generating a risk propagation network, and calculating a risk diffusion coefficient; performing hierarchical analysis on project requirements based on the risk diffusion coefficient to obtain a task decomposition scheme, constructing a task transmission chain based on the task decomposition scheme and generating a visual task relation graph, and generating a task state transition rule and an alternative task path based on an early warning state of a task node, meanwhile, constructing a task elastic scheduling mechanism and storing the task elastic scheduling mechanism into a rule database; and identifying a preorder task node and a post-order task node on the task transmission chain, sending the task reconstruction instruction to the preorder task node and the post-order task node to realize dynamic task adjustment, updating the risk propagation network according to an adjustment result, and optimizing the risk diffusion coefficient.

The invention provides a multi-modal large model task processing method, device and equipment based on reinforcement learning, and the method comprises the steps: employing a multi-modal large model, generating G groups of responses for multi-modal visual task data, carrying out the scoring of each group of responses, obtaining a reward value, carrying out the standardization of the reward value, obtaining a dominance score, building a strategy updating gradient through the dominance score, and carrying out the calculation of the strategy updating gradient. A final result is obtained after expectation reward maximization, model parameter adjustment and multi-round iteration, and an expectation reward objective function comprises an expectation calculation item and KL divergence constraint, so that a current strategy can be updated by comparing the performance of candidate strategies through group relative strategy optimization, and local optimum can be helped to be jumped out; the strategy updating amplitude is limited by the KL divergence constraint, and the model is prevented from violently changing in the optimization process, so that the training stability is improved; the dynamic strategy iteration allows the model to adjust the balance between exploration and utilization according to the learning progress on the basis of keeping the stability, thereby further ensuring the effectiveness and stability of strategy optimization.

The invention discloses a frequency domain enhancement fusion network method for night infrared and visible light images, and relates to the technical field of image processing. According to the method, a coarse enhancement network (CENet) is constructed, the night visible light feature extraction capability is improved through deep convolution and point-by-point convolution, a channel importance score is calculated in combination with SENet, finally, a pixel-level enhancement image is generated by using a Tanh activation function, the enhanced visible light feature image and an infrared light image are respectively sent to FDFNet, and the pixel-level enhancement image is obtained. In FDFNet, a U-shaped feature extractor is constructed based on EFA and D3FN modules of a frequency domain, visible light significant information is injected into an infrared feature map through SENet, so that the feature learning ability of a U-Net network is enhanced, finally, images passing through the FDFNet are spliced, a final fused image is obtained through image reconstruction, and the fusion effect is improved. The method provided by the invention is superior to the most advanced method in the aspect of LLVIP data set, shows the potential of further improving the fusion image precision in the future, and is helpful to promote the progress of advanced visual tasks.

The invention discloses a bionic moving target sensing, storage and calculation integrated photoelectric detection system based on continuous photoconduction, and relates to the field of bionic intelligent vision and the field of brain-like calculation, and the system comprises a synaptic bionic photoelectric detector array and a brain-like memristor array which are interconnected. When the synaptic-like bionic photoelectric detector array detects visual information of a moving target, a fixed bias voltage is applied to a device in each pixel, space illumination information corresponding to each pixel is converted into a continuously attenuated photocurrent signal, illumination information of a multi-frame dynamic target is continuously received in the continuous attenuation process of the photocurrent, and the visual information of the moving target is detected by the synaptic-like bionic photoelectric detector array. After the last frame of illumination information is input, a light current signal of each pixel is read, and a feature map memorizing all historical moment information is determined; the brain-like memristor array is based on the Ohm's law and the Kirchhoff's law, performs deep neural network operation on the feature map, outputs a current corresponding to a task result label, identifies a dynamic target, and can meet the requirements of a deep neural network and a complex dynamic visual task.

The invention belongs to the technical field of digital image processing, and discloses a low-light image enhancement method based on a zero-reference Retinex decomposition network, which comprises the following steps of: constructing a low-light image enhancement data set; building and training a low-light noise extraction network; building and training a low-light decomposition network, and estimating a reflection component and an illumination component; enhancing the low-light image; and testing the low light enhancement network. According to the low-light image enhancement method based on the zero-reference Retinex decomposition network, the illumination component, the reflection component and the noise component of the low-light image are learned through the deep learning network, the denoised low-light image is obtained, the illumination component after illumination adjustment is obtained through the gamma correction technology, and the low-light image enhancement method based on the zero-reference Retinex decomposition network is obtained through the Retinex decomposition theory. And obtaining a final low-light enhancement result. According to the method, training data is not needed, the low-light image is directly and rapidly enhanced, the image visual quality is improved, and the visual task precision is improved.

The invention discloses a visual algorithm self-training method based on multi-agent collaborative optimization, and the method comprises the following steps: constructing a multi-agent system architecture which comprises a user interaction layer, an intelligent scheduling layer, an A2A protocol communication layer and a professional agent cluster layer; the user interaction layer analyzes a user task intention and generates an execution plan; the scheduling agent calls the professional agent to complete data processing, model construction, training, testing and deployment; a task process is coordinated through a standardized communication mechanism, and task execution is supported by combining an MCP tool set, a knowledge base module and a memory system; and when the task fails, automatically executing rescheduling operation, and finally outputting a self-training result. According to the method, the development efficiency, the self-adaptability and the intelligent level are remarkably improved, and the method is suitable for computer vision tasks such as industrial detection, intelligent security and protection and automatic driving.

The invention discloses a Token-based visual task generation method, and belongs to the technical field of intelligent task automation, and the method comprises the steps: S1, cross-modal alignment; s2, performing visual Token processing; s3, constructing a task description Token sequence: constructing the task description Token sequence based on a predefined visual task template library according to requirements of a user or a specific application scene; s4, checking task feasibility; s5, task priority scheduling; s6, training a task generation model; s7, dynamic task allocation; and S8, optimizing the model. According to the method, the long sequence processing capability is optimized through a hierarchical merging strategy, the compactness of feature expression is realized while space position information is reserved, linear projection and enhanced position coding are combined to form a visual Token sequence with strong representation capability, local detail features are contained, a global context relationship is kept, and the method is suitable for the visual Token sequence with high representation capability. High-information-density feature input is provided for subsequent task processing, and the processing precision of various visual algorithms is effectively improved.

The invention discloses a pulse neural network training method based on adaptive threshold integer activation neurons, and the method specifically comprises the following steps: inputting a neuromorphic data stream or a static image, generating a pulse sequence through time coding, and taking the pulse sequence as the input of a model; in the backbone network, performing down-sampling and extracting multi-scale features by using an SNN module based on LIF neurons activated by an adaptive threshold integer; the visual tasks are divided into image recognition, image detection and image segmentation, and are respectively accessed to different processing modules for processing and outputting; according to the invention, on the basis of a Meta-SpikeFormer element module, the spiking neural network based on LIF neurons activated by an adaptive threshold integer value is used, so that the memory overhead during training is reduced, the network performance of direct training is improved, the application range of the spiking neural network is expanded, and the network is better applied to image classification, recognition and segmentation tasks.

The invention discloses a security and protection deployment and control management method and device. The method comprises the following steps: acquiring an initial deployment and control task parameter which is input by a target object in an interactive interface and comprises a deployment and control scene, determining extension prompt information matched with the initial deployment and control task parameter by utilizing a pre-trained computer vision large model, and displaying the extension prompt information in the interactive interface, obtaining a target deployment and control task parameter which is input by the target object in the interactive interface and obtained after the initial deployment and control task parameter is expanded according to the expansion prompt information, generating a deployment and control task according to the target deployment and control task parameter, and in the process of executing the deployment and control task, obtaining a monitoring image in real time, and comprehensively analyzing the monitoring image and the target deployment and control task parameters by using a computer vision large model to obtain a deployment and control task matching result corresponding to the monitoring image. According to the method, the technical problems that a related model algorithm in a security and protection deployment and control scene can only process a specific visual task and the intelligent degree is not high are solved.

The invention discloses an arbitrary-shape scene text detection method based on contour feature enhancement, and belongs to the technical field of computer vision. The invention provides an arbitrary-shape scene text detection method based on contour feature enhancement, and aims to solve the problems of edge contour missing and error detection in text detection so as to accurately position a text instance in a scene image. The method comprises the following steps: firstly, combining swin transform with a convolutional neural network, enhancing the perception of the network on a text region by utilizing rich detail information in a low-level feature map, and compensating for the missing of an edge contour; in addition, according to the method, the text area and the text-like area are distinguished through gradient information enhancement of the text edge, and therefore error detection is avoided. According to the method, a research scheme is provided for a deep learning method for natural scene text detection, and development of visual tasks such as image understanding is promoted to a certain extent.

The invention belongs to the technical field of artificial intelligence and computer vision, particularly relates to a visual processing edge adaptive decision-making method for resource and task perception, and aims to solve the problem of scheduling mismatch caused by resource dynamic change and task demand diversity in visual task processing in an edge computing environment. The method comprises the following steps: collecting multi-dimensional resource state data of edge nodes in real time to form a resource state vector with high time resolution; analyzing the visual task request, and constructing a quantifiable task feature vector; and establishing a resource-task association mapping model based on a dynamic weight distribution mechanism. The method also supports cross-edge domain collaborative decision, and processes a pipeline dynamic reconstruction and security isolation mechanism. According to the technical scheme, the fluctuation of the resource utilization rate is reduced to 15% or below, the average task processing delay is reduced to 60%, the scheduling satisfaction degree is improved by 40% or above, and the self-adaptability and the service quality guarantee capability of the edge vision system are remarkably enhanced.

The invention provides a visual token pruning method based on graph information propagation. The visual token pruning method comprises the following steps: performing visual extraction on an input image to obtain a visual token; initializing the importance score of the visual token; constructing a graph structure related to the visual tokens, taking each visual token as a node, calculating an adjacent matrix for constructing connection between the visual tokens, and initializing the graph structure; the adjacency matrix is updated through a preset similarity threshold value, and visual token sub-graph structures of different areas are obtained; performing normalization processing on each row of the adjacent matrix, performing iterative propagation on node information, and calculating a final score of each visual token; screening out k visual tokens with the highest scores according to the final scores of the visual tokens for projection; and splicing the visual token obtained by projection with the text token, and inputting the spliced visual token and text token into a large language model to obtain an output result. According to the method, the calculation efficiency of the model can be improved, and the calculation cost is remarkably reduced while the visual task performance is kept.

The invention discloses a stereo image quality evaluation method of a multi-task coder-decoder based on Transform, and belongs to the technical field of image quality evaluation. The invention provides a multi-task coding-decoding device network architecture based on Transform. The multi-task coding-decoding device network architecture takes image depth estimation as an auxiliary task; a feedback mechanism of a human visual system can be better simulated by adopting a coder-decoder network, the problems existing in left and right view feature extraction are solved, and an information processing process from a global state to a local state is realized; in addition, the global interaction capability of Transform and the local attention capability of CNN are combined to enhance image features, depth estimation is introduced as an auxiliary task, depth prior information with higher quality can be provided, cross-view parallax fusion based on Transform is guided, the problem existing in left and right view feature fusion is solved, and more accurate binocular features are obtained. And feedback compensation is carried out on multi-level binocular feature fusion by using an attention mechanism to obtain a final pixel-level quality map, so that the image quality is accurately evaluated, a research scheme is provided for a deep learning method for non-reference stereo image quality evaluation, and the development of a visual task is effectively promoted.

The invention provides a road environment perception method and a perception model training method, and relates to the technical field of image processing, in particular to the fields of automatic driving, computer vision, deep learning and neural network models. The specific implementation scheme of the road environment sensing method comprises the following steps of: obtaining image features from a look-around image of a vehicle by utilizing a feature extraction network of a first sensing model; wherein the look-around image is obtained according to a plurality of image acquisition devices at different angles; determining a bird's-eye view feature map according to the image features and the bird's-eye view query vector by using a coding network of a first perception model; and respectively determining output results of the plurality of perception tasks according to the bird's-eye view feature map by using the plurality of head networks of the first perception model. According to the technical scheme provided by the invention, computing resources consumed by simultaneous execution of multiple visual tasks can be effectively reduced, and the real-time performance and task effects of the multiple visual tasks are ensured.

The invention provides an image quality evaluation method and system based on multi-vision task fusion, and relates to the technical field of image processing. According to the method, the fitness of different images to a specific visual task is quantified into corresponding evaluation index values, and subjective scores of a user are replaced with the evaluation index values so as to guide training of an image quality prediction model. Specifically, the method comprises the following steps: firstly, constructing a multi-task-oriented image set used for image quality evaluation, and based on the influence of each image in the image set on the performance of different visual tasks (target detection, semantic segmentation and attitude estimation), quantifying to obtain evaluation index values of the different visual tasks on each image in the image set; the method comprises the following steps: firstly, acquiring an image set, fusing to obtain a fusion evaluation index value of each image in the image set, and then training based on the images carrying the fusion evaluation index values to obtain an end-to-end image quality prediction model which can be used for image quality evaluation.

The invention discloses an isovariant consistency image fusion method and system based on task driving, and the method comprises the steps: carrying out the feature decomposition of an input infrared image and a visible image, obtaining the basic features and detail features, and carrying out the geometric transformation operation meeting the isovariant consistency constraint in the subsequent fusion processing, so as to guarantee the transformation stability of the features. And performing semantic constraint on the fusion model by utilizing supervision information of a semantic segmentation task of the semantic segmentation model. According to the method, the equal-variant consistency constraint and the semantic loss are collaboratively optimized through the loss function part, the robustness of geometric transformation can be kept, meanwhile, the reservation of local details and the enhancement of global semantic information are effectively considered, a high-quality fused image is generated, and the performance of a downstream advanced visual task is improved.

The invention discloses a pulse neural network image classification method and system based on a space-time adaptive threshold, mainly solving the problems of poor nonlinear expression and limited time sequence and space feature processing ability in the prior art, and the scheme comprises the following steps: obtaining an image data set, and dividing the image data set into a training set and a test set; a spiking neural network main body structure comprising an input layer, a hidden layer and an output layer is selected, an existing neuron model is improved by introducing a space-time joint threshold adjustment mechanism, and improved neurons are placed in each neuron layer in the hidden layer to form a spiking neural network based on a space-time adaptive threshold. The training set is used to carry out iterative training; and inputting the test set into the trained pulse neural network to obtain an image classification result. According to the method, a space-time adaptive threshold mechanism is introduced, the threshold can be dynamically adjusted to adapt to time and space features, the processing capacity of the network on time sequence data and complex features and the classification accuracy of images are remarkably improved, and the method can be widely applied to dynamic visual tasks and event-driven scenes.

The invention belongs to the field of computer vision image signal processing, and relates to a multi-robot collaborative indoor scene semantic segmentation method and system based on multivariate interactive learning. The method comprises the steps of adopting an image feature encoder based on a double attention integration module to realize cross-modal feature fusion on an RGB image and a depth image; a dimension anisotropy attention enhancement module is adopted, and feature optimization is carried out through a multi-dimensional direction attention mechanism; a cross-task collaborative interaction module is adopted, and information is shared among different tasks by utilizing an interaction mechanism based on a feature level; in a multi-task heterogeneous decoding stage, each robot uses a decoder with a specific task to execute a specific task, and semantic segmentation, instance segmentation and scene classification are realized; and a joint optimization loss function is adopted to perform end-to-end joint training on multiple tasks, so that overall optimization and collaborative awareness performance improvement are realized. According to the method, the precision and robustness can be improved when complex indoor scene semantic segmentation and other visual tasks are executed.

The invention discloses a visual large model Token adaptive optimization method, system and device based on differential evolution and a medium, and the method comprises the steps: carrying out the data processing of an image classification data set, an instance segmentation data set and a saliency target detection data set, and obtaining all Tokens corresponding to each image through a Patch Embedding and position coding method; obtaining a plurality of groups of Tokens corresponding to each image through a random selection mode, and performing data processing to output all Tokens corresponding to each image and the plurality of groups of Tokens selected from each image; constructing a Token adaptive selection module, a self-attention optimization module and a downstream task output module; a complete Token adaptive optimization visual large model is constructed; training a reconstruction model and a complete Token self-adaptive optimized visual large model; performing model reasoning to obtain an image classification result, an instance segmentation result image and a saliency target detection result image; the system, the equipment and the medium are used for implementing the method. The method can be widely applied to various visual tasks such as image classification, instance segmentation and saliency target detection.

The invention discloses a continuous attention neural feedback training method and system based on a brain-computer interface, and relates to the technical field of neural feedback, and the method comprises the steps: collecting a multi-channel electroencephalogram signal of a user in visual task training in real time; extracting power spectral density characteristics of the multi-channel electroencephalogram signals in a beta frequency band, classifying the power spectral density characteristics by adopting a support vector machine algorithm, and outputting a judgment result of an alert or non-alert state; and according to a judgment result, dynamically adjusting an information fusion proportion alpha value in the visual task through a reward-punishment mechanism, updating image information feedback in the visual task in real time, and adjusting the attention state of the user through an image information feedback result. Neural feedback and a dynamic reward and punishment system are fused, real-time excitation feedback is obtained by autonomously adjusting electroencephalogram activity, the problem of insufficient training power caused by traditional static tasks or single positive feedback is solved, and the long-term training effect is enhanced.

The invention relates to the technical field of rock mass fracture recognition, in particular to a rock mass fracture recognition method based on a boundary detection neural network model. According to the method, an innovative deep learning framework is provided for boundary recognition, the characteristics of a refinement strategy and sub-pixel convolution are combined, and a leading result is obtained on a data set. The research shows that the CED model can improve the performance of middle-level and high-level visual tasks including optical flow estimation, target suggestion generation, semantic segmentation and the like. According to the method, a novel architecture is provided to solve the precision problem of a CED algorithm, a reverse refining path is supplemented for an HED network to improve the recognition precision, and the path uses efficient sub-pixel convolution to merge features step by step; the CED provided by the invention can generate an edge image which is well aligned with an image boundary.

The present disclosure provides a vision processing and model training method, device, storage medium and program product. A specific implementation solution is as follows: establishing an image classification network with the same backbone network as the vision model, performing a self-monitoring training on the image classification network by using an unlabeled first data set; initializing a weight of a backbone network of the vision model according to a weight of a backbone network of the trained image classification network to obtain a pre-training model, the structure of the pre-training model being consistent with that of the vision model, and optimize the weight of the backbone network by using real data set in a current computer vision task scenario, so as to be more suitable for the current computer vision task; then, training the pre-training model by using a labeled second data set to obtain a trained vision model.

The invention provides an unmanned aerial vehicle visual inertial odometer optimization method and device, equipment and a medium, and the method comprises the steps: carrying out image enhancement and denoising preprocessing on an image based on image information collected by an RGB-D camera of an unmanned aerial vehicle; constructing an image pyramid for the preprocessed image for subsequent multi-scale feature extraction; extracting optical flow change characteristics of two adjacent frames of images by using positive and reflective flow methods, determining a motion relation between adjacent image frames, and eliminating mismatching points by using an RANSAC algorithm; and according to a screening result, based on a motion relation between adjacent image frames, pose estimation is completed through loopback detection and a rear end, and then the pose of the camera is optimized according to the estimated pose to obtain an optimal pose estimation result of the visual inertial odometer. According to the method, the unmanned aerial vehicle can effectively execute the visual inertial odometer and the visual SLAM task in the complex environment, and the robustness positioning of the unmanned aerial vehicle in the complex environment is realized.

The invention discloses a multi-degraded image recovery method based on frequency domain decomposition, and aims to solve the problems that a single model is difficult to deal with various image degradation and recovery processes of different frequency domains are mutually coupled in the prior art. According to the method, a degraded image is decomposed into a high-frequency space and a low-frequency space through fast Fourier transform, and a double-branch network architecture is adopted for targeted processing: for the high-frequency part, a high-frequency feature adaptive processing module HFPM is designed, and detail texture features are effectively extracted and interference is suppressed through feature enhancement and cross-layer fusion technologies; and for the low-frequency part, constructing a low-frequency feature conversion enhancement module LTEM, and capturing global context information by using cyclic convolution to improve the integrity of the structure contour. According to the method, decoupling processing of frequency domain features is realized, and the image restoration performance of the model in various degradation scenes such as rain removal, noise removal and defogging is remarkably improved through the synergistic effect of high-frequency detail enhancement and low-frequency structure optimization. Experimental results show that the method has excellent recovery effect and robustness when a plurality of image degradation tasks are processed at the same time, and can be effectively applied to visual tasks such as traffic accidents with high image quality requirements.

The invention discloses a long-distance rail identification and obstacle detection method and system, and the method comprises the steps: combining a long-focus optical system with an ultra-high-speed industrial camera at a hardware level, and building a high-collection-frame-rate and long-distance optical detection system; on the software level, a region of interest for obstacle detection is delimited according to recognition and visual segmentation of left and right rail lines of a rail, advanced visual tasks such as a rail obstacle intrusion detection algorithm based on target detection are executed, and finally high-frame-rate recognition and detection of long-distance rails and rail obstacles are achieved. According to the invention, a Schmidt-Cassegrain optical system with a focal length of 2000mm is used, and an ultra-high-speed industrial camera with a resolution of 1920 * 1080 and 3000 frames per second is combined to build a long-distance high-frame-rate optical detection platform which is erected on a driving position of a train to detect railway conditions within a range of 6km in real time; serious accidents caused by overlong braking distance due to high-speed running of the train can be prevented. Compared with a traditional train driver eye observation and laser radar-based scanning imaging mode, a farther detection distance, better detection and recognition robustness and a higher frame rate can be realized through the long-distance optical observation platform and the rear-end real-time detection system.

The invention provides a visual task processing method and device based on a multi-modal large model of low-rank adaptive enhancement, and relates to the technical field of data processing. The method comprises the following steps: receiving a plurality of visual task requests sent by a visual application and visual input data corresponding to each visual task request in the plurality of visual task requests; dynamically scheduling the plurality of visual task requests into a plurality of batches of processing sequences through an orchestrator, and determining a reasoning mode of a low-rank adapter used by each batch of visual input data; and inputting the visual input data of each batch and the corresponding low-rank adapter into a target multi-modal large model, and performing dynamic fragmentation calculation on the visual input data and a decomposition matrix of the corresponding low-rank adapter through an adaptive fragmentation matrix multiplication operator ATMM to obtain a response result corresponding to each visual task request. Through the method provided by the invention, diversified visual task requirements are met.

The invention discloses an infrared and visible light image fusion network system and an application method thereof, and relates to the technical field of image fusion. According to the method, a feature extraction block is constructed preferentially, and a multi-scale separable convolution module and deformable convolution are combined in the feature extraction block, so that extraction of environment information and highlighted information in a fused feature map is improved; then, by introducing a space channel cooperation attention module, channel information and space information of the feature map are refined, meanwhile, the channel information and the space information are made to interact to aggregate more information, and the information capturing capacity of the feature map is improved; experimental results show that the performance of the proposed method on an MSRS data set is superior to that of the most advanced method, and the future potential research direction is highlighted, so that the precision of the fused image is improved, and meanwhile, the development of advanced visual tasks is promoted.

The invention relates to an image enhancement method suitable for an underground coal mine low-illumination image. The method comprises the steps that a low-illumination target image is loaded into a low-illumination image enhancement model, image enhancement processing is carried out on the low-illumination target image through the low-illumination image enhancement model, and the low-illumination target image is decomposed into a corresponding target image illumination image and a target image reflection image; reconstructing and generating a target image enhanced illumination image; performing denoising processing on the target image reflection image generated by decomposition so as to generate a denoised reflection image of the target image after denoising processing; and carrying out merging processing on the target image enhanced illumination image and the target image denoised reflection image to generate a target enhanced image after merging processing. According to the method, the low-illumination image of the underground coal mine can be effectively enhanced, the requirement for computing resources is reduced, the reasoning process of image enhancement is accelerated, and the visual task requirement of the underground coal mine is met.

The invention discloses an infrared and visible light image fusion method based on target perception, and the method comprises the following steps: S1, obtaining a training data set which comprises a plurality of input image pairs composed of infrared images and visible light images; s2, constructing an image fusion depth model; s3, training the image fusion depth model based on the training data set to obtain a target image fusion depth model; and S4, fusing the to-be-fused image pair by using the target image fusion depth model to obtain a fused image result. According to the invention, through the target sensing region fusion sub-model, the space-frequency attention sub-model and the fusion block sub-model, differential feature extraction and fusion are realized, and the definition, detail retention and downstream visual task performance of the fused image are improved.

The invention discloses a semantic matching method based on visual features of a large-scale pre-training model, and the method comprises the following steps: obtaining an image data set which comprises a target image and a reference image with corresponding semantics; constructing a corresponding semantic matching network based on the visual basic model framework, wherein the semantic matching network comprises a feature extraction module, a staggered sensing module, a matching module and an up-sampling module; training a semantic matching network by using the image data set to obtain a semantic matching model for image semantic matching; and inputting the reference image, the to-be-matched points on the reference image and the target image into a semantic matching model to output matched points on the target image. According to the method provided by the invention, the accuracy of the semantic matching task can be effectively improved, and a better service is provided for a downstream visual task.